Notice: Community Accounts Have Changed. What You Need To Know >>

Topic - aliased halo around label decal

Page 1 of 2
   1  2  Next  Last 
See the second image for issues I'm having.
I never have found the solution for this despite a couple of threads that have discussed this issue. it's bloody inconvenient and hard to fix in post.
In this Thread my issue was dark halo around white colors. now it's the opposite.
I thought i'd take another stab at posting this out there in case someone can help. look especially and the dark gray flourishes against the dark bottle. I've tried both transparent png and Tiff file with separate alpha mask. same result. Texture filtering makes little difference.

www.rodsg.com
http://www.flickr.com/photos/rodsawatsky/

Message edited by Rod Sawatsky on 3/21/2013 - 5:36 PM

another issue I find is with render passes. for some reason the shadow density pass typically comes out looking quite unlike the actual shadows in the scene. Intensities are quite off and don't seem to respect the relative strengths of the shading in the scene. Does anyone else out there find the shadow density pass of any use?

www.rodsg.com
http://www.flickr.com/photos/rodsawatsky/

If nobody has a solution to this but has a workflow in post to fix things like this would you care to share?

currently I am production a mask for the decal separately and then contracting the mask in photoshop by 1 or 2 pixels to lose the edge but it can kill a bit of detail sometimes.

ideas?
thanks

www.rodsg.com
http://www.flickr.com/photos/rodsawatsky/

The colour in the RGB channel simply needs to bleed outside the limits of the alpha.

Alternatively, I'd be tempted to simply create separate greyscale masks for the text and the swirl and use them in conjunction with two 'constant' layers in the shader tree.
Thanks for piping up Ian. I think I recall trying the "bleed" thing a while back to no avail and indeed this seems to be the case again here. and it looks pretty much identical whether using separate tiff files for diffuse color and alpha or one single transparent PNG file.

the separate alphas with constant colors may be a workaround in this particular case but is really a very specific solution to this instance and wouldn't help me in scenarios where color is more complex (halftones). I'd dearly love to find a fix that works a bit more broadly.

thanks so much for posting your ideas.

www.rodsg.com
http://www.flickr.com/photos/rodsawatsky/

Message edited by Rod Sawatsky on 3/27/2013 - 3:47 PM

I've tried recreating the problem, but even with the texture on a white background (i.e. no bleed) it renders ok. Without knowing more details of your shader tree I'm a bit stumped. Could you post a screengrab? Or maybe even create a similar scene using generic geometry and textures?

One question: What are your antialiasing settings for the images? They should be OFF.
I had all kinds of troubles until i switched to TGA/Targa.. Dunno but maybe give it a whirl
Ok, I've posted up a sample file with my actual decal supplied as a tiff with separate alpha and as a png file. let me know if you find a cleaner solution.

Link Here

www.rodsg.com
http://www.flickr.com/photos/rodsawatsky/

Just to be clear regarding the previous couple posts, the antialiasing is always off and I have tried the TGA file format as well. none of these basic things seem to affect any change. can someone confirm or test this from the provided file?

is there anything in 701 that fixes this rendering issue?

www.rodsg.com
http://www.flickr.com/photos/rodsawatsky/

Well, I just bit the bullet and upgraded to 701 last night. I had planned to wait this one out and skip until 801 but just couldn't handle the psychological effect of "looking over the fence" and wondering...primarily the speed enhancements.

I pulled up the scene in 701 and hit render with no changes to settings and unfortunately the problem persists with no differences.

anyone else?

www.rodsg.com
http://www.flickr.com/photos/rodsawatsky/

The main problem appears to be a bug with "group mask". It changes the diffuse amount of layers underneath it.

The fix is to pull "Cherish Shiraz label" above the base shader, give it it's own shader, then LAYER MASK the shader instead.

NOTE: Don't use "group mask" here in 701. You will run into the same bug as before. In 601 group masking the shader seems to work.

Second problem is your diffuse color image should have some bleeding. In this case it's not very noticeable because your bottle is black, and the tif background is black. If your bottle was white and you zoomed in, you would see a faint black halo around the logo.

I have attached a fixed image with some bleed. In PS I did as follows:
1) I took your tif, cut the layer out using the alpha as a selection
2) Layer > matting > remove black matte
3) Used xnormal > dilation, to bleed it out 16 pixels (you can use other methods to bleed)

--
Win7 64/Core i7-975 3.33GHz/12GB RAM/GTX 670 2GB (344.11)/2x1920x1200/Wacom Intuos4 M
I am the resurrection, and the life: he that believeth in me, though he were dead, yet shall he live - Jesus Christ

Message edited by funk on 4/15/2013 - 11:29 AM

I got rid of a fringe that was bugging me by changing the gamma of the group mask
Funk, thanks for dropping in here and giving your detailed input. for some reason I'm not having any results with the masked shader approach. same problem. I haven't tried bleeding the art because the art doesn't seem to need it. the red areas aren't giving me a problem and the dark gray flourishes are butting up to black in the background so if anything the edges should be dark and not light...at least if bleeding is the issue. Your closeup seems to be resolving the issue somehow though. would it be possible to post your updated file to look at? can't see what I'm doing differently.

Ripper...thanks for that last tip about the gamma on the mask. that one's a winner and actually has the best results for me so far and is far simpler to adjust than anything else. I set the gamma to .1 and it rendered pretty clean. It would never have occurred to me and I can't quite understand why that works like that. Can anyone from Luxology comment on why this behaves this way?

...baffled but pleased.

www.rodsg.com
http://www.flickr.com/photos/rodsawatsky/

Message edited by Rod Sawatsky on 4/16/2013 - 9:47 PM

Actually, on further testing it appears that a combination of both is needed here. group mask isn't working well with this technique of lowering gamma so I tried the layer mask on a shader in the material group and set it's gamma to .1 and it worked quite well.

I really hope this helps others in their tests because it has been a pain in the backside on a number of projects and is a big relief to find a workaround. I hope it holds up to testing in other scenarios.

www.rodsg.com
http://www.flickr.com/photos/rodsawatsky/

Here is the fixed scene. Select FILE > Download to download the zip file

--
Win7 64/Core i7-975 3.33GHz/12GB RAM/GTX 670 2GB (344.11)/2x1920x1200/Wacom Intuos4 M
I am the resurrection, and the life: he that believeth in me, though he were dead, yet shall he live - Jesus Christ

Message edited by funk on 4/17/2013 - 1:18 PM

thanks for the file Funk. I confirm your findings that group mask works in 601 but not in 701 (with this file). It seems we're running into a number of comments about things that have reverted back a step since 601 and I hope these little details get resolved.

my tests with your files confirm my suspicion that the bleeding effect seems to have no impact on the render results. I believer there may be cases where this tactic is useful though. My experience in the past with white type having a dark halo was not helped by this approach though so I'm unsure when it would really make a difference.

your input was indispensable here so thanks for chiming in and sharing your file.

www.rodsg.com
http://www.flickr.com/photos/rodsawatsky/

you know, talking about taking steps back in 701, I'm finding render speeds on these reflective bottles to be consistently slower than 601. Any of you finding the same?

www.rodsg.com
http://www.flickr.com/photos/rodsawatsky/

Quote from Rod Sawatsky :
thanks for the file Funk. I confirm your findings that group mask works in 601 but not in 701 (with this file). It seems we're running into a number of comments about things that have reverted back a step since 601 and I hope these little details get resolved.



Although we have found a minor bug in 701 (which I reported), I want to clear some things up.

1) The original approach you were using doesnt work in any version of modo, because it's actually the wrong way to do it. When you mix 2 materials, it HAS to be done by masking SHADERS, otherwise the top material values will over-ride some of the lower material's values in unwanted ways. It may "appear to sort of work" in some cases but that's only because you got lucky and had similar values in both materials. It's also buggy in 601 and 701.

2) Using GROUP MASKS always renders much slower than layer masks, I assume because modo has to process everything in the group first to create the mask. With a layer mask, it only uses ONE layer. I always avoid group masks where possible.

So the final solution, using layer masks to control the group shader, is actually the best way to do it. If I was creating your scene from scratch I would have done it this way and never noticed other methods were buggy.

Quote from Rod Sawatsky :

my tests with your files confirm my suspicion that the bleeding effect seems to have no impact on the render results. I believer there may be cases where this tactic is useful though. My experience in the past with white type having a dark halo was not helped by this approach though so I'm unsure when it would really make a difference.



I tested your original files on a white bottle and could see a slightly dark halo. It was minor but I'm a perfectionist, so I created a new version with bleeding. Personally I think you should always do this because it takes less than 10 seconds to create.

I dont think you can rely on your past results, because as I stated earlier, you have been doing it the wrong way.

--
Win7 64/Core i7-975 3.33GHz/12GB RAM/GTX 670 2GB (344.11)/2x1920x1200/Wacom Intuos4 M
I am the resurrection, and the life: he that believeth in me, though he were dead, yet shall he live - Jesus Christ

Message edited by funk on 4/18/2013 - 8:47 AM

Well, Funk, this last post of yours is definitely informative but it further underscores the prevailing confusion people seem to have with the shader tree. You say it is the WRONG way to set up the material and yet it is exactly the way James Darknell describes in this video. He mentions having separate material controls for the decal etc. as I would expect but mentions nothing about the need for a unique shader. I have seen similar tuts or instructions for material setup on this forum without mention of a unique shader needed.

While experience and practice favors your statement currently it has clearly been far from obvious in the past as I actually sent this same related problem to Luxology's tech support last year(in this thead) with no successful advice from Lux or the community. I actually fixed it in Photoshop at the time as it wasn't worth the headache anymore. So after all that it sure feels like something isn't being communicated clearly in the training or docs when you say it is actually the wrong way to do it....

Do you mind me asking where you found this explained clearly? Other people should know this too.

thanks also for the tip about rendering speed issues with Group Mask. I'll run some comparisons.

Rod.

www.rodsg.com
http://www.flickr.com/photos/rodsawatsky/

Message edited by Rod Sawatsky on 4/19/2013 - 9:07 PM

Quote from Rod Sawatsky :
Well, Funk, this last post of yours is definitely informative but it further underscores the prevailing confusion people seem to have with the shader tree. You say it is the WRONG way to set up the material and yet it is exactly the way James Darknell describes in this video.


Yeah I agree it can all be a bit confusing. It even seems to confuse some of the more knowledgeable modo users. As I said, masking materials can work in many cases, but sooner or later you are going to run into problems. James has probably never mixed 2 materials that were different enough to cause problems so he has never noticed.

Quote from Rod Sawatsky :

While experience and practice favors your statement currently it has clearly been far from obvious in the past as I actually sent this same related problem to Luxology's tech support last year(in this thead) with no successful advice from Lux or the community.


That's interesting... It wasn't James responding to your tech support email was it?

I noticed in that thread you posted, Will Trumble directed you to another thread where Superdimensional mentions the same solution I gave you (although he doesnt go into detail).


Quote from Rod Sawatsky :

I actually fixed it in Photoshop at the time as it wasn't worth the headache anymore. So after all that it sure feels like something isn't being communicated clearly in the training or docs when you say it is actually the wrong way to do it....

Do you mind me asking where you found this explained clearly? Other people should know this too.


I don't remember reading this in the docs (not saying this info isnt there)

I think I figured this out while trying to create car paints which have multiple layers (clearcoat, flakes, paint). Some posts by captain obvious helped too.

Once you start doing complex layered materials you quickly realise that this kind of stuff can only be done using masked shaders.

--
Win7 64/Core i7-975 3.33GHz/12GB RAM/GTX 670 2GB (344.11)/2x1920x1200/Wacom Intuos4 M
I am the resurrection, and the life: he that believeth in me, though he were dead, yet shall he live - Jesus Christ

I recall going through that other thread Funk...but they never mentioned the Layer mask on the shader and how Group mask seemed to have some bug. I never found a solution consequently and find your explanation here the clearest I've seen so far. I'm in your debt and hope it is a help to others finding the same confusion.

www.rodsg.com
http://www.flickr.com/photos/rodsawatsky/

I thought I'd add a newly discovered anomaly with this otherwise perfect solution that funk has so clearly explained. I'm adding it here so that this continues to add depth as a resource. In this context with the dark wine bottle everything is fine. however, I tried this same setup on a clear plastic bottle and the label decal ended up adding unusual shadow density to the ground plain in a way you wouldn't expect.

the top version shows the way I used to have it set up with the plastic bottle material and label decal below the base shader...the shadow cast by bottle and label looks as I'd expect.

the bottom version shows the shader tree stacked as I've learned to do in this thread but the shadow cast with the label material on is much darker. If I turn the label material off it looks more like the top version. Somehow it doesn't seem to be calculating the transparent areas correctly for the shadow...looks like the shadow of a solid colored object.

can someone confirm this behavior?

www.rodsg.com
http://www.flickr.com/photos/rodsawatsky/

Message edited by Rod Sawatsky on 5/16/2013 - 6:29 PM

Man...adding a second shader on top of this for the label really inflates render times a lot. Is there not a better way?

Why do you need the second shader unless you're trying to do some sort of tricky material blending or need refinement control? Sorry just trying to understand here for this example.

Rod- I get the same issue if I try and do things this way too- it's like the glass is now made of a solid and the shadows rays don't make through the glass.

Message edited by Gideon Klindt on 6/27/2013 - 4:36 PM

And just when I thought I had finally found a workaround to fix all problems! Ecchh. I'm glad to hear you say it doesn't make sense Gideon cuz I sure didn't think it should be necessary to have a separate shader...and then using a layer mask on the shader instead of group mask...not straightforward at all.

still I'm glad for the pointers here. adds to the troubleshooting library. a better fix that improves render times would definitely be welcome.

www.rodsg.com
http://www.flickr.com/photos/rodsawatsky/

Yeah, I'm not saying you don't have to do it sometimes...but in this case, for your instance I'm lost as to why you'd have or want to make it two different shader and make everything get evaluated so many extra times. I could be completely wrong on this count, but I'm trying to see the value in this case.

The fringe issue with these decals IS a problem with MODO IMHO, unless somebody can outline a sure fire way to do it proper and with only one/two files. The bleed method works, but it says to me that something else is wrong and needs to either be addressed in the manual, or fixed in the software.

Message edited by Gideon Klindt on 6/27/2013 - 6:02 PM

Page 1 of 2
   1  2  Next  Last